Enhanced oil recovery methods classification using radial basis function neural network

A radial basis function neural network is presented to classify enchanced oil recovery (EOR) methods using reservoir and fluid parameters. The methodology is similar to that of Surguchev and Li (2000), but a different strategy is used to group EOR methods into classes. The methodology allows a fast assessment of applicability of EOR methods with limited available reservoir information. We used as input data twelve parameters associated to eighteen EOR methods, which represent the output data and are grouped by number of methods into eleven classes. The network is trained and validated using 330 and 94 patterns, respectively. After the training process, the network is considered satisfactory for assessing the applicability of EOR methods if the network can generate for all validation patterns at least one target method of the class to which the patterns belong. The best result shows that the network is able to classify 90% of the validation patterns related to the different classes.